Abstract: The vibrating screen is an important equipment in coal washing process. The amount of materials on its screening surface has a direct impact on production efficiency and management level. However, the traditional manual monitoring method has problems such as large subjective errors, inability to conduct real-time monitoring, and high labor intensity. The volume of materials on the screening surface of the vibrating screen is small with indistinct features, making the classic object detection algorithms incapable of meeting the detection requirements. To tackle the above problems, a deep learning-based algorithm MFI-YOLOv7 for detecting amount of materials on vibrating screen′s surface is proposed. With YOLOv7 as the basis, MFI-YOLOv7 adopts the Omni-Dimensional Dynamic Convolution (ODConv) in the Backbone network to enhance the feature extraction ability of the network. In the Neck layer, a CARAFE-FPN feature fusion structure is designed to strengthen feature fusion. In the Prediction layer, a prediction box loss function, Focal-CIOU Loss, is designed to enhance the positioning ability of the network. To verify the detection effect, the effectiveness of MFI-YOLOv7 is verified through ablation experiments and comparative experiments. The results of the ablation experiments show that the introduction of ODConv, CARAFE-FPN and Focal-CIOU Loss has improved the precision, recall rate and the mean average precision of the model with these three indicators being increased by 1.68, 1.07 and 1.68 percentage points, respectively. The results of the comparative experiments indicate that MFI-YOLOv7 outperforms classic object detection algorithms such as Faster R-CNN, SSD, CenterNet, YOLOv7 and YOLOv10 in terms of both detection accuracy and speed. In addition, the application effect of MFI-YOLOv7 in actual scenarios was also tested and analyzed. The results show that the detection effect of this algorithm is superior to other algorithms under different lighting conditions and at different camera positions, and material states. In scenarios with large changes in lighting conditions, the material detection rate of MFI-YOLOv7 is seen to increase by 5 to 21 percentage points. In the case of detecting concentrated and loosely scattered the materials, the detection rates of MFI-YOLOv7 are 82.67% and 68.75% respectively. The camera position also has an impact on the detection effect. The MFI-YOLOv7 proposed offers an effective solution for the automatic detection of the amount of materials on vibrating screen surface, and is helpful to improvement of the production efficiency and intelligent management level of coal preparation plants.